IBM SPSS Modeler Cookbook

IBM SPSS Modeler Cookbook

Overview of this book

IBM SPSS Modeler is a data mining workbench that enables you to explore data, identify important relationships that you can leverage, and build predictive models quickly allowing your organization to base its decisions on hard data not hunches or guesswork. IBM SPSS Modeler Cookbook takes you beyond the basics and shares the tips, the timesavers, and the workarounds that experts use to increase productivity and extract maximum value from data. The authors of this book are among the very best of these exponents, gurus who, in their brilliant and imaginative use of the tool, have pushed back the boundaries of applied analytics. By reading this book, you are learning from practitioners who have helped define the state of the art. Follow the industry standard data mining process, gaining new skills at each stage, from loading data to integrating results into everyday business practices. Get a handle on the most efficient ways of extracting data from your own sources, preparing it for exploration and modeling. Master the best methods for building models that will perform well in the workplace. Go beyond the basics and get the full power of your data mining workbench with this practical guide.

IBM SPSS Modeler Cookbook

Credits

Foreword

About the Authors

About the Reviewers

www.PacktPub.com

Preface

Free Chapter

Data Understanding

Introduction

Using an empty aggregate to evaluate sample size

Evaluating the need to sample from the initial data

Using CHAID stumps when interviewing an SME

Using a single cluster K-means as an alternative to anomaly detection

Using an @NULL multiple Derive to explore missing data

Creating an Outlier report to give to SMEs

Detecting potential model instability early using the Partition node and Feature Selection node

Data Preparation – Select

Introduction

Using the Feature Selection node creatively to remove or decapitate perfect predictors

Running a Statistics node on anti-join to evaluate the potential missing data

Evaluating the use of sampling for speed

Removing redundant variables using correlation matrices

Selecting variables using the CHAID Modeling node

Selecting variables using the Means node

Selecting variables using single-antecedent Association Rules

Data Preparation – Clean

Introduction

Binning scale variables to address missing data

Using a full data model/partial data model approach to address missing data

Imputing in-stream mean or median

Imputing missing values randomly from uniform or normal distributions

Using random imputation to match a variable's distribution

Searching for similar records using a Neural Network for inexact matching

Using neuro-fuzzy searching to find similar names

Producing longer Soundex codes

Data Preparation – Construct

Introduction

Building transformations with multiple Derive nodes

Calculating and comparing conversion rates

Grouping categorical values

Transforming high skew and kurtosis variables with a multiple Derive node

Creating flag variables for aggregation

Using Association Rules for interaction detection/feature creation

Creating time-aligned cohorts

Data Preparation – Integrate and Format

Introduction

Speeding up merge with caching and optimization settings

Merging a lookup table

Shuffle-down (nonstandard aggregation)

Cartesian product merge using key-less merge by key

Multiplying out using Cartesian product merge, user source, and derive dummy

Changing large numbers of variable names without scripting

Parsing nonstandard dates

Parsing and performing a conversion on a complex stream

Sequence processing

Selecting and Building a Model

Introduction

Evaluating balancing with Auto Classifier

Building models with and without outliers

Using Neural Network for Feature Selection

Creating a bootstrap sample

Creating bagged logistic regression models

Using KNN to match similar cases

Using Auto Classifier to tune models

Next-Best-Offer for large datasets

Modeling – Assessment, Evaluation, Deployment, and Monitoring

Introduction

How (and why) to validate as well as test

Using classification trees to explore the predictions of a Neural Network

Correcting a confusion matrix for an imbalanced target variable by incorporating priors

Using aggregate to write cluster centers to Excel for conditional formatting

Creating a classification tree financial summary using aggregate and an Excel Export node

Reformatting data for reporting with a Transpose node

Changing formatting of fields in a Table node

Combining generated filters

CLEM Scripting

Introduction

Building iterative Neural Network forecasts

Quantifying variable importance with Monte Carlo simulation

Implementing champion/challenger model management

Detecting outliers with the jackknife method

Optimizing K-means cluster solutions

Automating time series forecasts

Automating HTML reports and graphs

Rolling your own modeling algorithm – Weibull analysis

Business Understanding

Introduction

Define business objectives by Tom Khabaza

Assessing the situation by Meta Brown

Translating your business objective into a data mining objective by Dean Abbott

Produce a project plan – ensuring a realistic timeline by Keith McCormick

Index

Customer Reviews

5 star

4 star

3 star

2 star

1 star

Using an empty aggregate to evaluate sample size

Having all the data made available is usually not a challenge to the data miner—the challenge is having enough of the right data. The data needs to be relevant to the business question, and be from an appropriate time period. Many users of Modeler might not realize that an Aggregate node can be useful even when all you have done is drag it into place, but have given no further instruction to Modeler.

At times data preparation requires the number of records in a dataset to be a data item that is to be used in further calculations. This recipe shows how to use the Aggregate node with no aggregation key and no aggregation operations to produce this count, and how to merge this count into every record using a Cartesian product so that it is available for further calculations.

Getting ready

This recipe uses the cup98lrn reduced vars2 empty.txt data set. Since this recipe produces a fairly simple stream, we will build the stream from scratch.

How to do it...

To use an empty Aggregate node to evaluate sample size:

Place a new Var. File source node on the canvas of a new stream. The file name is cup98lrn reduced vars2.txt. Confirm that the data is being accessed properly.
Add both an Aggregate node and a Table node downstream of the source. You do not need to edit either of the nodes.
Run the stream and confirm the result. Total sample size is 95412.
Now, add a Type node and a Distinct node in between the Source and Aggregate node. Move the variable CUST_ID into the Key fields for grouping box.
Select Discard only the first record in each group.
Run it and confirm that the result is 0. You have learned that there are no duplicates at the customer level.
Place a Merge node so that it is combining the original source with the output of an empty Aggregate.
Within the Merge node choose Full Outer Join.
You have just successfully added the total sample size to the data set where it can be used for further calculation, as needed.

How it works...

What an Aggregate node typically does is use a categorical variable to define a new row—always a reduction in the number of rows. Scale variables can be in the Aggregate field's area and summary statistics are calculated. Average sales in columns arranged with regions in rows would be a typical example. Having given none of these instructions, the Aggregate node boils our data down to a single row. Having given it no summary statistics to report all, what it does is the default instructions, namely Include record count in field, which is checked off at the bottom of the Aggregate node's menu. While this recipe is quite easy, this default behavior is sometimes surprising to new users.

There's more...

Now let's talk about some other options, or possibly some pieces of general information that are relevant to this task.

If you are merging many sources of data, as will often be the case, you should check sample size for each source, and for the combined sources as well. If you obtained the data from a colleague, you should be able to confirm that the sample size and the absence (or presence) of duplicate IDs was consistent with expectations.

When duplicates are present, and you therefore get a non-zero count, you can remove the aggregate and export the duplicates. You will get the second row (or third, or even more) of each duplicate. You can look up those IDs and verify that they should (or should not) be in the data set.

A modified version

A modified version of this technique can be helpful when you have a nominal variable with lots of categories such as STATE. Simply make the variable your key field.

Additionally, it is wise to sort on Record_Count with a Sort node (not shown). The results show us that California has enough donors that we might be able to compare California to other states, but the data in New England is thin. Perhaps we need to group those states into a broader region variable.

The same issue can arise in other data sets with any variable of this kind, such as product category, or sales district, etc. In some cases, you may conclude that certain categories are out of the scope of the analysis. That is not likely in this instance, but there are times when you conclude that certain categories are so poorly represented that they warrant a separate analysis. Only the business problem can guide you; this is merely a method for determining what raw material you have to work with.

IBM SPSS Modeler Cookbook

IBM SPSS Modeler Cookbook

Overview of this book

Related Content you might be interested in

Current Title:

IBM SPSS Modeler Cookbook

Using an empty aggregate to evaluate sample size

Getting ready

How to do it...

How it works...

There's more...

A modified version

See also